Evidence scan:Reducingprescribing errorsApril 2012Identify Innovate Demonstrate Encourage
ContentsKey messages	                                                                                              31.	Sco...
Key messagesMedicines can do a lot of good but they also have the potential tocause harm. Medication errors are one of the...
1.	ScopeHealth professionals and managers are always looking for ways toimprove the quality and safety of healthcare. Medi...
A study to develop a definition of prescribing errors    Human factors approaches address the interactionsin the UK conclu...
1.3	Methods                                               While the focus is on prescribing errors, rather                ...
2.	 Education and developmentTraining personnel to better prepare them for tasks and workconditions is a human factors app...
prescribing error on the exam, but all passed on the     Another hospital in Spain tested ways to improvesecond attempt an...
2.2	 Reducing errors                                    identified and randomly assigned to provider                      ...
Error monitoring and reportingMonitoring and reporting on errors may in itselfserve to raise awareness and support improve...
3.	 Expanding professional rolesSelecting appropriate personnel is an important human factorsapproach. This section descri...
Checking medication orders                             errors was lower than before the intervention andA number of studie...
who alerted specialists to prescribing errors, gave     medication errors to the pharmacist than to theirpatients informat...
3.3	 Other humanfactors issuesOther human factors issues such as fatigue,concentration levels and stress may all have anim...
4.	ToolsRedesigning equipment and tasks can reduce prescribing errors.This section describes 80 studies about tools that h...
E-prescribing systems aim to reduce delay in             in a nephrology outpatient clinic at a paediatricaccessing medica...
In Spain, a hospital unit using handwritten              Most research focuses on the potential ofprescriptions was compar...
E-prescribing systems have sometimes been               abbreviation errors. However, errors not associatedassociated with...
But when three primary care clinics in the US          It may be that decision support is more useful atimplemented e-pres...
Prescribing excessive doses is a common                Similarly, researchers in the US tested whether aprescription error...
Other researchers in the US examined three             4.3	 Human factors issuespersonal digital assistant (PDA)-based dru...
Three sites in the US implementing e-prescribing       Types of alertsidentified barriers, including those relating to    ...
Similarly, researchers in the US aimed to improve        4.4	Standardisedclinician acceptance of drug alerts in 31 primary...
randomised to receive 12 evidence-based drug           receives from different hospitals nationwide.therapy letters immedi...
5.	Summary5.1	 Key points                                          errors, particularly if the systems do not allow the   ...
Reducing prescribing errors   evidence scan-2012
Reducing prescribing errors   evidence scan-2012
Reducing prescribing errors   evidence scan-2012
Reducing prescribing errors   evidence scan-2012
Reducing prescribing errors   evidence scan-2012
Reducing prescribing errors   evidence scan-2012
Reducing prescribing errors   evidence scan-2012
Reducing prescribing errors   evidence scan-2012
Reducing prescribing errors   evidence scan-2012
Reducing prescribing errors   evidence scan-2012
Reducing prescribing errors   evidence scan-2012
Upcoming SlideShare
Loading in...5

Reducing prescribing errors evidence scan-2012


Published on

Published in: Health & Medicine
  • Be the first to comment

No Downloads
Total Views
On Slideshare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Reducing prescribing errors evidence scan-2012

  1. 1. Evidence scan:Reducingprescribing errorsApril 2012Identify Innovate Demonstrate Encourage
  2. 2. ContentsKey messages 31. Scope 42. Education and development 73. Expanding professional roles 114. Tools 155. Summary 25References 29 Health Foundation evidence scans provide information to help those involved in improving the quality of healthcare understand what research is available on particular topics. Evidence scans provide a rapid collation of empirical research about a topic relevant to the Health Foundations work. Although all of the evidence is sourced and compiled systematically, they are not systematic reviews. They do not seek to summarise theoretical literature or to explore in any depth the concepts covered by the scan or those arising from it. This evidence scan was prepared by The Evidence Centre on behalf of the Health Foundation.© 2012 Health Foundation
  3. 3. Key messagesMedicines can do a lot of good but they also have the potential tocause harm. Medication errors are one of the most common causesof patient harm and prescribing accounts for a large proportion ofmedication errors. This evidence scan examines strategies to reduceprescribing errors.Prescribing errors include mistakes or inaccuracies Educational strategieswhen choosing and ordering treatments, such as Educational initiatives tend to focus on stoppingwrong doses or illegible prescriptions. errors before they occur. Strategies include:Eight databases were searched and 123 studies were –– group training sessionsincluded about strategies for reducing prescribingerrors, predominantly from North America. Studies –– individual education visitsabout errors of omission, such as not prescribing –– letters and printed materialsa drug that might be helpful, were excludedbecause it is difficult to be objective about what –– audit and error reporting systemsmedications should be prescribed in any individual –– improvement projects and collaboratives.instance. The scan does not cover other medication All of these initiatives have had some success, buterrors such as those related to dispensing or there is not enough evidence to say which strategiesadministration. work best.Most studies about reducing prescribing errorshave been undertaken in hospital. The three most Professional rolescommonly researched approaches are, in order of Studies of expanding professional roles tend tofrequency: computerised tools, training to improve focus on how pharmacists can identify any errorsprescribing and expanding professional roles to before patients are harmed, including:identify errors. –– checking for errors as prescriptions are receivedComputerised tools at the pharmacy or on wardsElectronic prescribing and computerised decision –– medicine reconciliation or reviewssupport have been studied extensively but there are –– individual or group education sessions.mixed findings. Most studies suggest computerisedtools can reduce prescribing errors but some Most research suggests that engaging pharmacistssuggest unintended negative consequences. in these ways can be beneficial, but few studies haveEmerging evidence suggests that to be successful, explored the best ways to integrate pharmacistshuman factors such as workflow features, tool into teams and the interprofessional factors todesign and context need to be considered. be considered. Combining education, enhanced professional roles and computerised tools may help to reduce prescribing errors most effectively.THE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 3
  4. 4. 1. ScopeHealth professionals and managers are always looking for ways toimprove the quality and safety of healthcare. Medicines are a keycomponent of healthcare and errors relating to medication mayimpact on patient safety. This evidence scan explores what is beingdone to reduce prescribing errors.1.1 Purpose 1.2 DefinitionsThousands of people in the UK take medicinesevery day to help manage ongoing conditions Prescribing errorsor to help them through an emergency or crisis. Prescribing is the process whereby a doctor,Most receive and take their prescriptions without nurse or other registered professional authorisesincident in hospital or in the community, but in a use of medications or treatments for a patientsmall number of cases an error occurs, whether or and provides instructions about how and whennot it is evident to patients. those treatments should be used. Although the term commonly refers to orders for medicines, A medication error is a failure in the the concept can equally encompass laboratory treatment process that leads to, or has the tests, medical imaging, psychological treatments, potential to lead to, harm to the patient.1 eye glasses, eating and exercise regimes or other instructions to help optimise health andThis evidence scan explores steps that have been wellbeing.2,3explored to minimise prescribing errors. It does notcover the frequency or cause of prescribing errors. Prescriptions are handwritten or computerisedIt focuses solely on approaches that have been used documents containing the patient’s name andto minimise such errors. address, the date, the specific treatments prescribed and an authorising signature. They are a way forThe scan addresses the questions: prescribers to communicate with pharmacists or– What approaches have been used to reduce others who in turn fill the prescription. Prescribers prescribing errors? include doctors of various types and, in some countries, nurse practitioners, physicians assistants,– Have any approaches related to human factors dentists, podiatrists, optometrists, clinical been researched? psychologists and clinical pharmacists also writeThe scan provides a rapid collation of empirical prescriptions.4–6research about initiatives to reduce prescribingerrors. All of the evidence has been sourced and Prescriptions can help people stay healthy orcompiled systematically, but the scan is not a manage long-term conditions or emergencysystematic review and does not seek to summarise situations. However, as with other components ofevery study on this topic. healthcare, prescriptions are also subject to error and can lead to unintended harm. MedicationThis section defines prescribing errors and human errors are one of the most common patient safetyfactors approaches and describes the methods used issues and prescribing errors are one of the mostto identify relevant research. The following sections common types of medication errors.7–12outline the three broad approaches that have beenused to reduce prescribing errors: training to avoid Prescribing errors can take many forms, butprescribing errors before they happen, expanding commonly involve incorrect doses, illegible detailsprofessional roles to identify and rectify errors, and or ordering inappropriate medications or drugsusing tools to improve processes. that may react with other medications already being taken.THE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 4
  5. 5. A study to develop a definition of prescribing errors Human factors approaches address the interactionsin the UK concluded that transcription errors, between people, the work environment andfailure to communicate essential information organisational systems. This discipline seeksand the use of drugs or doses inappropriate for to understand people’s psychological andthe individual patient were prescribing errors, physiological limitations, the demands imposedbut omissions and deviations from policies or upon people at work and how the mismatchguidelines were not.13 Even so, some also define between the two leads to errors.prescribing omissions as errors, for example if adoctor fails to prescribe an antihypertensive drug In the field of healthcare, human factors approachesfor someone who could benefit from it. aim to enhance clinical performance through an understanding of the effects of teamwork, tasks,In this evidence scan, the focus is on active errors, equipment, workspace, culture and organisationwhereby the prescription contains a potentially on human behaviour and abilities and to apply thatharmful drug, combination or dosage rather than knowledge in clinical settings.solely errors of omission. Studies focused onlyon errors of omission were excluded unless they Human factors approaches may involveexplicitly defined such errors as ‘prescribing errors’ diagnosing issues in the interaction betweenand investigated definite strategies to reduce those people and systems, identifying workloaderrors. In this way, the scan uses the definition of and task interruptions and redesigning the‘prescribing errors’ as outlined in individual studies workplace environment and team factors throughin the review. This means that studies that sought standardisation and prioritisation.to ‘improve prescribing’ in terms of adherence to Human factors approaches tend to focus onguidelines or increasing or decreasing the rates personnel, training and operating parameters.14–21of prescribing some types of medicines were not More specifically, human factors solutions mayincluded unless the authors specifically defined include five broad approaches:22,23these as prescribing errors. The scan focuses onresearch about reducing prescribing errors rather –– training individuals to better prepare them forthan research about ‘improving prescribing’ more the work and conditionsgenerally. –– selecting individuals who possess the best characteristics for the job and avoiding fatigue,Human factors stress and burnoutAll approaches to reduce prescribing errors were –– environmental design such as improvedof interest but there was a special focus on human lighting, temperature control and reduced noisefactors approaches. –– equipment design including tools and‘Human factors’ is a multidisciplinary field automationincorporating contributions from psychology, –– task design to change what staff do rather thanengineering, design, ergonomics, operations just the devices they use. This may involveresearch, aviation, continuous quality improvement assigning some or all of tasks to other workers orand other disciplines. to automated processesIn general terms, a ‘human factor’ is a physical, Interventions to reduce prescribing errors allmental, emotional or social aspect that is specific fit broadly into these human factors categories.to humans and may influence how people interact There is potential overlap in these categories butwith the environment and people around them. interventions to reduce prescribing errors haveHuman factors approaches thus study all aspects of focused largely on training individuals, selectingthe way humans relate to the world around them, individuals (pharmacist roles) and equipment andtheir capabilities and limitations and how these can task design (including electronic systems). Thebe used to improve performance and safety. following chapters describe research about each of these three areas in turn.THE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 5
  6. 6. 1.3 Methods While the focus is on prescribing errors, rather than medication errors more generally, theThe scan focused on research articles published review screened studies related to ‘medicationin journals in the UK and internationally and was errors’ because often studies use this terminologycompleted over a two-week period. rather than the term ‘prescribing’. Studies usingTo identify relevant research, one reviewer searched this broader terminology were included if theyeight bibliographic databases for studies of any contained specific data about prescribing errors.design in any language published between 1990 More than 10,000 articles were scanned and 123and early January 2012. The databases comprised studies met the inclusion criteria.Medline, Embase, the Cochrane Library andControlled Trials Register, PsychLit, Google Although prescribing of all types was eligible forScholar, Web of Science, ScienceDirect, and the inclusion, only research about reducing errors inHealth Management Information Consortium. medication prescribing was identified in the search. The report therefore focuses on reducing errors inSearch terms included combinations of prescribing medication prescribing.error, prescription error, medication error, dosingerror, dose error, human factors, task identification, Findings were extracted from all publications usingtask redesign, workplace environment, situational a structured template and studies were groupedawareness, team roles, standardisation, according to key themes to provide a narrativeprioritisation, workload interruptions, pharmacist, summary of trends.pharmacy, computerised order entry, computerisedphysician order entry, computerised pharmacistorder entry, e-prescribing and similes.General search terms such as ‘prescribing error’were used first to identify the largest range ofstudies. When research about specific interventionswas identified, such as e-prescribing, theseinterventions were then added as search terms toensure completeness. In this way the search strategywas initially general, to identify research about awide variety of interventions, and then becamemore detailed, to gain in-depth information aboutthe specific interventions identified.To be eligible for inclusion, studies had to be readilyavailable empirical research or systematic literaturereviews which examined some type of outcomerelating to reducing prescribing errors. This mayinclude, for example, strategies used to reduceerrors, the benefits or costs of doing so, or increasesin reporting rates. Studies about the number andtype of prescribing errors were not included.Studies that described potential approaches toreducing prescribing errors but did not containempirical data were also excluded.THE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 6
  7. 7. 2. Education and developmentTraining personnel to better prepare them for tasks and workconditions is a human factors approach. This section describes24 studies about training and educational initiatives to reduceprescribing errors.Research has focused on using training and Another review of interventions to improvedevelopment initiatives to reduce prescribing errors prescribing included 29 studies. Educationalin two distinct ways. outreach visits and audit and feedback were most commonly studied and were found to be effective–– The first relates to reducing errors during the for improving prescribing practice.25 prescribing process itself. Here, research has examined one-to-one and group education and Researchers in Australia tested the value of improvement projects to stop errors happening academic detailing to reduce simple errors when in the first place. prescribing drugs that can be addictive in hospital.–– Second, research has examined training and One hospital acted as a control and another development initiatives to identify and rectify received academic detailing, where junior doctors any errors that do occur, to minimise the chance received an educational visit and a bookmark of them harming patients. Here, the focus tends reminder containing the requirements for selected to be on error monitoring and reporting systems. drugs. Prescription error rates decreased from 41% to 24% at the hospital receiving academic detailingResearch about training and development and the confidence of junior doctors in writinginitiatives is divided into these two subsections prescriptions increased. There was no change inbelow. error rates at the control hospital.26 Elsewhere in Australia, a hospital used a decision2.1 Reducing errors support tool to help with drug dosing for peopleduring prescribing with kidney problems. The system was introduced to prescribers using academic detailing. There wereOne-to-one education improvements in dosing for various drugs. TheIndividualised education can take many forms, evaluators concluded that one-to-one educationincluding ‘academic detailing’ whereby a helped to introduce tools for reducing prescribingprofessional is visited in their workplace for a errors.27one-to-one education session. A Cochrane Reviewexamined face-to-face outreach visits by a trained Group education for traineesperson to a health professional. 18 randomised Most studies of group education sessions to reducetrials were included, 13 of which targeted prescribing errors focus on training for medical orprescribing. All outreach interventions included pharmacy students or registrars.several components such as written materials orconferences. Reminders or audit and feedback A prescribing skills course for interns was testedwere sometimes used. All studies found improved in the US. A pharmacy faculty gave two lectures,behaviours. However, few studies examined patient attended hospital rounds and took part in clinics.outcomes or costs.24 Interns then undertook a written exam and clinical assessment. All interns made at least oneTHE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 7
  8. 8. prescribing error on the exam, but all passed on the Another hospital in Spain tested ways to improvesecond attempt and gained prescribing privileges handwritten prescriptions in neonatal units. Staffafter six months. The researchers concluded that took part in training about good prescribingthe prescribing curriculum was practical and practice and used a pocket PC automatic dosagefeasible.28 However, studies like these tend not calculation system. Incorrect prescriptions reducedto follow up on results to examine the impact on from 40% to 12%.32reducing prescribing errors in practice. Other studies have examined combined trainingAn Objective Structured Clinical Examination for both fully qualified professionals and trainees.(OSCE) is an assessment tool that uses lay people Researchers in England examined whethertrained to respond to questions in a standardised prescriber education in tutorials, ward-basedmanner. Students’ performance is observed teaching and feedback with each new group ofand scored. An OSCE station related to the trainee medical staff could reduce prescribingcommunication and management of prescription errors in intensive care. Prescribing audits beforeerrors was tested with third-year students at one training, immediately after training and six weeksUS medical school. In total, 77% of students said after training were fed back to prescribers withthat the OSCE station improved their awareness of their individual prescribing and error rates andmedication errors and 71% thought that they were anonymised information about other prescribers’more comfortable communicating prescription error rates. Prescription errors decreased.33errors to patients. Feedback about root causeanalysis, collaboration with the pharmacist for Improvement programmeserror analysis, interpersonal and communication A small number of studies have tested howskills feedback from faculty, use of a standardised collaborative improvement projects and networkspatient and use of an actual prescription that led to of professionals may impact on prescribing errors.a medication error were thought to be helpful.29 Thirteen hospitals in one US state took part in aBut not all types of training for students are collaborative project to improve medication safety.successful. A hospital in Canada tested a 30-minute Teams were encouraged to make changes to theirtutorial for all fellows and residents starting in the medication processes based on evaluating theiraccident and emergency (A&E) department at the medication systems and ergonomic principles andbeginning of the academic year. The tutorial was research. Before and after data from eight of thefollowed by a written test. Prescribing errors were hospitals suggested a 27% decrease in medicationreviewed on 18 randomly selected days. There was errors, a 13% increase in error detection andno difference in prescribing error rates between prevention and a 24% increase in formal writtenthose who attended and those who did not attend reporting of errors that reached the patient.34the tutorial.30 A hospital in Argentina implemented strategies toGroup education for change the safety culture and reduce medication errors in children and babies. Interventionsqualified professionals focused on promoting positive safety cultureEducation to reduce prescribing errors has also without punitive management of errors andtargeted fully qualified professionals. A neonatal specific prescribing and drug administrationintensive care unit in Spain evaluated the effect of recommendations. The medication error rateeducational sessions for health professionals on decreased from 11% to 7% over a two-year period.the number and type of prescription errors. The Prescribing errors were not analysed separately butprescription error rate reduced from 21% to 3%.31 were specifically targeted.35THE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 8
  9. 9. 2.2 Reducing errors identified and randomly assigned to provider feedback or usual care. However, after one yearafter prescribing there was no difference in adverse drug events.40One-to-one educationVarious types of individualised education have Group education for traineesalso been studied for reducing the impact of errors Researchers in Canada evaluated a computeror identifying errors before they harm patients. training module to improve third-year pharmacyIn Australia, direct feedback to clinicians was students’ ability to identify and correct prescribingtested to reduce errors from polypharmacy or errors. The module helped increase thedrug interactions in older people. GPs were sent identification of errors.41information about the at-risk patient, relevant In the US, first-year pharmacy students took part inclinical guidelines and a personalised covering laboratory simulations to help identify and preventletter. There was a reduction in the average medication errors, including prescribing errors.number of medications prescribed for each person Following simulations and role plays, students’following the prescriber feedback.36 knowledge and awareness of medication errorsSimilarly, researchers in Canada examined whether improved as did their confidence in recognising andfollow-up letters from pharmacists to doctors preventing errors and communicating about them.42following inappropriate prescriptions would However, studies like these tend not to follow up toimprove prescribing for people in long-term care. examine the impact on reducing prescribing errorsThe educational letters briefly described potentially in practice.inappropriate prescriptions and suggestedalternatives. 38% of potentially inappropriateprescriptions were changed by the doctor following Improvement programmes A hospital in Switzerland tested various approachesa letter.37 for reducing the impact of adverse drug incidents.Researchers in the US tested whether a Non punitive incident monitoring was set up in acomputerised drug review database linked to a neonatal-paediatric intensive care unit. Systemstelepharmacy intervention reduced inappropriate changes included double checking for potentiallymedication use in 23,269 people aged 65 years or harmful drugs, using a standardised prescriptionolder. Computer alerts triggered telephone calls to form and contacting the national drug controldoctors from pharmacists with training in older agency about misleading drug labels. Most ofpeople’s medicine who could discuss substitution the system changes were based on minor criticaloptions. As a result, 24% changed to a more incidents which were only detected after a longappropriate drug.38 period of time. They resulted in some potential errors being caught, including prescribing errors.43Education may also be informal and result frominteractions between staff members. Researchers in Another popular educational and improvementthe US assessed the views of pharmacy directors, method is audit and feedback. Changes are monitoredmedical centre executives and pharmacists over time and prescribers and pharmacists are givenabout the value of pharmacist residency training written feedback about their own performance inprogrammes. Participants believed that residency comparison to others. A Cochrane Review aboutprogrammes had many benefits and that these audit and feedback included 37 randomised trials,outweighed costs. They thought that pharmacy including some about prescribing. Effects wereresidents helped to reduce medication errors by varied. The reviewers concluded that audit andeducating prescribers and checking prescribing.39 feedback can sometimes be effective in improving the practice of health professionals, in particularPatients have been targeted for education in a prescribing and diagnostic test ordering, but effectssmall number of instances. In one study, 913 US tend to be small.44outpatients with potential prescribing errors wereTHE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 9
  10. 10. Error monitoring and reportingMonitoring and reporting on errors may in itselfserve to raise awareness and support improvement.A neonatal intensive care unit in Spain testedwhether prescribing errors would decrease merelyas a result of observation and recording of errors.The prescription error rate reduced from 33% to19%. Rates of incorrect dosing and lack of dosespecification in prescriptions reduced significantlybut there was no change in transcription errors.45A hospital in New Zealand conducted auditsover a 10-year period to improve the quality ofwritten prescriptions. Initially there was a highrate of insufficient documentation and illegibleprescriptions. Interventions designed to addressdeficiencies included feeding back audit results,education sessions for doctors and nurses onprescribing and medication errors and changesto systems, such as modifying medication charts,developing hospital-wide prescribing standards andan alert notification system. Over time, legibilityand documentation improved.46A system was set up to report on medication errorsat one hospital in France. 60% of medication errorsrelated to prescribing. The system was found tobe feasible and resulted in steps being taken toreduce errors. Success factors included a blame-free approach and ensuring that the system wasconfidential.47THE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 10
  11. 11. 3. Expanding professional rolesSelecting appropriate personnel is an important human factorsapproach. This section describes 19 studies about initiatives relatingto roles and personnel to reduce prescribing errors.As with studies about education and development, errors decreased from 10% to 5% and the numberresearch has focused on using varying professional of warnings that doctors complied with increasedroles and skill mix to reduce prescribing errors in from 44% to 68%.48two distinct ways.–– The first relates to roles during the prescribing 3.2 Reducing errors process to reduce the likelihood of errors after prescribing happening in the first place. Here, research is very limited, and has examined prescribing by Pharmacist roles nurses versus doctors. Most studies about reducing errors after–– Second, research has examined the use of prescriptions have been written have been healthcare professionals to identify and rectify undertaken in hospital, particularly in the US. any errors that do occur, to minimise the chance The most common interventions related to specific of them harming patients. Here, the focus tends roles focusing on pharmacists. to be on expanding the role of pharmacists to Pharmacist roles to identify prescribing errors and perform checks and identify errors. to stop them reaching patients include:Research about roles is divided into these twosubsections below. –– checking for errors as prescriptions are received at the pharmacy and contacting prescribers for clarification or amendment before filling3.1 Reducing errors prescriptionsduring prescribing –– visiting wards to review charts and provideFew studies have examined how professional advice to prescribers about individual patientsroles can be expanded to reduce prescribing –– reconciling the medicines patients usually takeerrors. One exception is a study of engaging with what they are prescribed in hospitalnurses in roles usually performed by doctors. Ahospital in Iran tested whether a collaborative –– providing medication reviews upon discharge.prescription order entry method consisting of Each of these initiatives is explored in turn.nurse order entry followed by doctor verificationand countersignature is as effective as a strictly Pharmacists have also run one-to-one or groupphysician order entry method in reducing education sessions for prescribers but theseprescribing errors in the neonatal ward. In both interventions tend to focus on prevention rathersystems a warning and suggested change appeared than error identification. Studies of this nature werewhen the dose or frequency of the prescribed covered in the previous section.medication was incorrect. The rate of medicationerrors was 40% lower for nurse order entrycompared to doctor order entry. PrescriptionTHE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 11
  12. 12. Checking medication orders errors was lower than before the intervention andA number of studies have examined the value of preventable adverse drug events were reduced. Theasking pharmacists to specifically check and review intervention cost 3 Euro per monitored day butmedication orders. For instance, pharmacists at a potentially saved 26 to 40 Euro per monitored dayUS hospital used an electronic system to review by preventing adverse drug events.54all prescriptions. This alerted the prescriber and Similarly, pharmacists reviewed prescriptionspharmacist to dosage errors and allergies and on the surgical wards at one hospital in Canadareduced prescription errors.49 and provided group educational sessions forAnother US hospital examined how paediatric doctors. Doctors accepted 90% of pharmacistclinical pharmacists intercept prescription errors. recommendations. There was a 9% decrease in drugIn total, 78% of potentially harmful prescribing costs.55errors were intercepted by pharmacists.50A hospital in England examined the impact of Medicine reconciliation Medication reconciliation, whereby a pharmacistpharmacists on preventing prescribing errors at checks usual medicines against planneddischarge. Routinely collected data showed that prescribing, can take place in hospital or in primary8% of all medication orders had an intervention care. A systematic review of four studies examinedby a pharmacist. Pharmacists intercepted 83% of medication reconciliation interventions in botherroneous orders without referring to doctors. settings. One randomised trial and one before andOmission, drug selection and dosage errors were after study evaluated pharmacist medication reviewthe most common.51 at hospital discharge. Neither found a benefit.Researchers in the Netherlands analysed the costs Two before and after studies examining systematicand benefits of hospital pharmacy staff detecting medication reconciliation at each primary care visitprescribing errors. Over a five-day period, 10% of had conflicting findings.563,540 medication orders in two Dutch hospitals In the UK, a pharmacist independent prescribercontained an error. Estimated benefits amounted completed systematic medicine reconciliation into 9,867 Euro compared to 285 Euro in staff time A&E and initiated an inpatient prescription chart.costs.52 Medicine reconciliation completed within 24 hoursBut interventions involving checks by pharmacists of admission increased from 50% to 100% andare not always successful. Researchers in France prescription chart initiation in A&E increased fromexamined how pharmacy validation can be 6% to 80%. The prescribing error rate was reducedused as a secondary filter for eliminating errors from 3.3 errors to 0.04 errors per patient.57from a computerised order entry system. All Elsewhere in the UK, a cost analysis of five differentprescriptions over a five-day period were analysed strategies for preventing medication errors atat one hospital. Pharmacy validation produced hospital admission used models and previousonly a moderate short-term impact on potential studies. Pharmacist reconciliation of medicines wasprescribing errors.53 found to be cost effective.58Pharmacists on wardsAnother strategy is to engage pharmacists to check Pharmacist discharge services One hospital in the Netherlands examined theprescribing on hospital wards. In the Netherlands, effect of a clinical pharmacist discharge service ona clinical pharmacist reviewed medication orders medication discrepancies and prescription errors infor patients admitted to the intensive care unit and people with heart failure. One group received usualdiscussed recommendations during patient review care by doctors and nurses. The other received ameetings with attending doctors. Over an eight review of discharge medication by pharmacistsand a half month period, the rate of prescribingTHE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 12
  13. 13. who alerted specialists to prescribing errors, gave medication errors to the pharmacist than to theirpatients information, prepared a written overview doctor. Pharmacists acted as the final interceptors,of discharge medication and communicated with detecting errors in prescriptions before theycommunity pharmacists and GPs. The pharmacist reached patients.62discharge service was associated with fewermedication discrepancies and prescription errors at Pharmacists may contact primary care doctorsone-month follow up (39% versus 68% of people in to clarify prescriptions or suggest changes. In thethe control group).59 US, call backs from pharmacies to 22 primary care practices were logged over a two-week period. Keeping records of the number and type of queriesMultifaceted hospital interventions from pharmacists helped practices develop specificSometimes a range of interventions involving interventions to reduce errors.63pharmacists are implemented simultaneously. Onepaediatric intensive care unit in Egypt introduced Some proactive approaches to pharmacist reviewa structured medication order chart, doctor have also been tested. In Switzerland six qualityeducation by pharmacists, provision of dosing circles were set up whereby six communityassists and performance feedback to doctors. pharmacists reviewed the prescribing of 24 GPs.Prescribing error rates reduced from 78% to 35%. Key elements included the review of specificPotentially severe errors reduced from 30% to 7%.60 prescriptions, continuous quality improvement and education, local networking and feedback ofA systematic review examined the frequency of comparative data about costs and drug choices.medication and prescribing errors in neonatal Analysis of nine years’ worth of data foundintensive care units. In 11 studies, the highest improved quality and safety of prescribing and areported rate was 5.5 medication errors per 100 42% decrease in drug costs compared to a controlprescriptions, but rates varied widely between group, representing savings of US$225,000 per GPstudies partly due to differences in definitions and per year.64methods. Dose errors were the most common.Computerised physician order entry, participationof pharmacists in ward rounds and pharmacist Nursing homesreview of prescriptions prior to dispensing were A review of 18 randomised trials of interventionssuggested to improve medication safety, but there to improve prescribing in nursing homeswere few high-quality evaluation data available.61 found that seven studies described educational approaches such as outreach visits, five studies described clinical pharmacist activities such asPharmacists in primary care medication reviews and two studies describedStudies are also available about the role of computerised decision support. Two studiespharmacists in reducing prescribing errors in described multidisciplinary approaches and twoprimary care. However, most of the research in this described multifaceted approaches. Improvementsarea is relatively small scale and descriptive and in prescribing were found in 83% of studies. Inobservational. It tends to describe interventions some cases, this included reductions in prescribingundertaken in a small number of sites, and little errors, but most of the interventions focused ondetailed or long-term data about outcomes are improving suboptimal prescribing which wasavailable. outside the scope of this evidence scan.65For instance, a US study examined the pharmacist’srole in improving medication safety in primary careusing focus groups with pharmacists and patients.Patients were likely to see multiple doctors but onlyone pharmacist. They were more likely to reportTHE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 13
  14. 14. 3.3 Other humanfactors issuesOther human factors issues such as fatigue,concentration levels and stress may all have animpact on prescribing behaviour. It has also beensuggested that temporary staff are more likely to beassociated with medication errors.66While descriptive articles are available aboutthese human factors concepts and their potentialimpact on safety issues, no empirical researchwas identified about interventions targeting thesefactors specifically to reduce prescribing errors.THE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 14
  15. 15. 4. ToolsRedesigning equipment and tasks can reduce prescribing errors.This section describes 80 studies about tools that have been used toreduce prescribing errors.Human factors approaches are concerned with Similarly, electronic prescribing is already standardthe interface between tools and systems and the in primary care in the UK, whereas in the US thispersonnel responsible for them. The majority is just beginning to get established. A great dealof studies about re-designing equipment and of research has been undertaken in the US abouttasks to reduce prescribing errors focus on e-prescribing systems, but the findings perhapselectronic prescribing systems (e-prescribing) merely serve to reinforce what is already standardand computerised decision support systems. practice in the UK.These studies tend to describe implementation ofspecific systems and their outcomes, but do notusually examine the interlinkages with workflow, 4.1 E-prescribinginterruptions and other human factors. Hospital careStudies about computerised tools are described in E-prescribing is also known by the termsthis section because the majority of research about computerised physician order entry (CPOE),reducing prescribing errors has focused on such computerised provider order entry ortools. However, it is acknowledged that the research computerised pharmacist order entry (in the UStends to focus on the technology rather than the where pharmacists may transcribe prescribers’interface between technology and personnel. handwritten orders into a computer system). This is an electronic process for entering instructionsAlmost all of the studies focus on how tools can about patient treatment. Orders for medication,be used to reduce errors during the prescribing equipment or other treatments are communicatedprocess itself, but some of the tools can also be over a computer network to various medical staffused as a way of identifying errors after they have and departments such as pharmacy, laboratory oroccurred. radiology who are, in turn, responsible for filling those orders.When interpreting the findings in this section it isimportant to remember that there are differences Before e-prescribing systems were available, in thein prescribing and in the roles of pharmacists in US doctors traditionally wrote out or verbally statedvarious countries. For example, electronic systems their instructions for patient care, which were thenhave been set up to reduce transcription errors but transcribed by nurses or ancillary staff before beingtranscription errors do not apply in the same way actioned. It was thought that such handwrittenin the UK as in the US. In the UK, doctors write notes may result in more errors and delays67 and, asdirectly onto a drug chart or into an electronic a result, the US Institute of Medicine recommendedprescribing system rather than onto a piece of e-prescribing be implemented as standard.68paper which is then transcribed by someone else.Studies that focus on reducing transcription errorsof this nature are therefore of limited relevance tothe UK.THE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 15
  16. 16. E-prescribing systems aim to reduce delay in in a nephrology outpatient clinic at a paediatricaccessing medication or treatment, reduce errors hospital. The overall prescribing error rate was 77%related to handwriting or transcription, allow for handwritten items and 5% with e-prescribing.orders to be made at the point of care or off-site and Before e-prescribing, 73% of items were missingsimplify inventory and charging processes. essential information and 12% were judged illegible. After e-prescribing was introduced, 1% ofThe systems often have decision support tools items were missing essential information and therebuilt in whereby the system automatically checks were no illegibility errors. The number of error-freefor duplicate or incorrect doses or tests, provides patient visits increased from 21% to 90%.73alerts to let the prescriber know that a dose istoo high or may interact with other medications, Researchers in Canada examined the impact ofor highlights clinical guidelines or other ways to e-prescribing on medication errors and adverseimprove evidence-based treatment. This section drug events in hospitalised children over a six-yearincludes studies about e-prescribing systems with period. Compared to wards using handwrittenand without inbuilt decision support tools (often orders, the computerised system was associatedthe distinction is not made clear in the studies). with a 40% lower medication error rate. However,The next subsection examines research about the there was no impact on adverse drug events.74impacts of decision support tools themselves. Over a four-year period, a US hospital introducedA large number of studies have found benefits from an e-prescribing system and incorporated decisione-prescribing, and it is commonly suggested that support features. The medication error ratesuch tools can reduce prescribing errors by around (excluding missed doses) fell by 81%. Seriousa half.69,70 medication errors that were not intercepted fell by 86%. Dose errors, frequency errors, route errors,For instance, a systematic review found that substitution errors and allergies all reduced.7523 out of 25 studies about e-prescribing whichreported on the medication error rate found Another US hospital implemented e-prescribingimprovements. Six out of nine studies that with features designed to improve medicationanalysed the effects on potential adverse events safety such as required fields, use of pick lists,found reduced risks. Four out of seven studies enhanced workflow features, alerts and remindersthat analysed the effect on actual adverse drug and access to online reference information. Theevents found reduced risks. Studies of locally system was associated with a reduced error rate.76developed systems, those comparing e-prescribingto handwritten prescriptions and studies using A US A&E department found that beforemanual chart review to detect errors, found greater e-prescribing there were 222 prescribing errorsimprovements.71 per 100 orders compared to 21 per 100 orders afterwards.77Another review of 12 studies comparedhandwritten versus computerised prescription Another study tested e-prescribing in a USorders. 80% of studies about e-prescribing children’s critical care unit. Before implementation,reported fewer prescribing errors compared with there were about 2 potential adverse drug eventshandwritten orders. The use of e-prescribing was per 100 orders compared to 1 per 100 ordersassociated with a 66% reduction in prescribing afterwards. There was a 96% reduction in errors.78errors in adults, but not children.72 A before and after study in a public hospital inStudies from many parts of the world with diverse Pakistan found that prescribing errors for inpatientshealth systems have found that e-prescribing were 23% during paper-based prescribing and 8%systems can reduce prescribing errors. For example, after the introduction of e-prescribing. The error rateresearchers in England assessed e-prescribing for patients upon discharge was 17% for paper-based prescribing and 4% after introducing e-prescribing.79THE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 16
  17. 17. In Spain, a hospital unit using handwritten Most research focuses on the potential ofprescriptions was compared with another using e-prescribing to avoid errors during the initiale-prescriptions. Handwritten prescriptions were prescribing process, but these tools can also beassociated with a 20% error rate compared to 9% used to identify errors after the prescription hasin electronically assisted prescriptions. Omission been entered. A US hospital aimed to reduceerrors were also lower with e-prescriptions.80 oral chemotherapy related prescribing errors intercepted by clinical pharmacists prior toIntensive care units at one hospital in Belgium reaching the patient. A multidisciplinary teamtested whether a computerised system could reduce identified key elements of the oral chemotherapythe incidence and severity of prescription errors. process using healthcare failure modes and effectsOne unit used a paper-based system and another analysis (HFMEA) then implemented e-prescribingused e-prescribing. There were fewer prescription which reduced the risk of prescribing error byerrors with the computerised system (3% versus 69%.86 Pharmacists used the system to check and27%) and fewer adverse drug events.81 amend prescriptions.A hospital in France compared two prescribing and E-prescribing systems have also been used to try tomedication distribution systems on a paediatric reduce errors indirectly. For example, researchersnephrology ward: a handwritten prescription in England tested whether data routinely producedplus ward stock distribution system versus by an e-prescribing system could be used tocomputerised prescription plus unit dose drug identify doctors at higher risk of making a seriousdispensing system. Over an eight-week period, the prescribing error, with the aim of intervening withcomputerised prescription error rate was 11% and these doctors. 848,678 prescriptions by 381 juniorthe handwritten prescription error rate was 88%.82 doctors at one hospital over a year long periodA hospital in the Netherlands tested decision were analysed. Doctors varied greatly in the extentsupport and computerised order entry. The to which they triggered and responded to alerts ofproportion of prescriptions containing one or more different types. It was not possible to use data abouterrors reduced from 55% to 17%.83 the number and type of alerts to identify doctors at high risk of making serious errors.87Some hospitals have modified or developedspecialised e-prescribing systems to target people Not all studies of e-prescribing have foundwith particular conditions or to address specific favourable results. Researchers in Canadatypes of errors. A systematic review of e-prescribing evaluated commercially available prescribingin hospital paediatric care and neonatal, paediatric software in hospital outpatient clinics. Data fromor adult intensive care settings included 12 22 weeks when the system was not availableobservational studies. Meta analysis found a were compared with 44 weeks when the systemdecreased risk of prescription errors. There was was available. During intervention weeks, aboutno reduction in adverse drug events or mortality 8% of prescriptions were electronic and the restrates.84 were handwritten. There was no difference in prescription error rates88 but this may be due to theDose calculation errors are the most common very low uptake rate of the system.type of medication error in children and babies.A systematic review examined interventions to A hospital in Portugal examined an e-prescribingreduce the risk of this type of error. 28 studies were system with a dose distribution tool. The toolincluded, mostly about e-prescribing. Most studies helped to reduce medication errors related toof e-prescribing found some reduction in errors. transcribing and patient identification, butHowever, one study found increased mortality after prescription and monitoring errors remained.89the implementation of e-prescribing.85THE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 17
  18. 18. E-prescribing systems have sometimes been abbreviation errors. However, errors not associatedassociated with negative or unexpected outcomes with abbreviations increased during the transitiontoo, including an increase in some types of errors.90 period.94For instance, a systematic review of 12 studiespublished between 1998 and 2007 examined A hospital in Italy compared manual prescriptione-prescribing in hospital. Nine studies found versus a computerised system. When thereduced prescribing error rates for all or some drug computerised system was first introduced thetypes, usually regarding minor errors. But several number of errors increased due to incompletestudies reported increases in the rate of duplicate dose and incomplete prescriptions. However, afterorders and failures to discontinue drugs. This the system was modified the overall rate of errorswas attributed to inappropriate selection from a decreased.95dropdown menu or not being able to view all active Some suggest that e-prescribing may takemedication orders concurrently. The reviewers longer than handwritten prescriptions.concluded that evidence for e-prescribing systems Researchers in England assessed a combinedis not compelling and is limited by small sample e-prescribing, automated dispensing, barcodesizes and poor study designs.91 patient identification and electronic medicationResearchers in the US examined hospital staffs’ administration record system in a hospital surgicalinteraction with an e-prescribing system at one ward. Prescribing errors reduced from about 4% tohospital over a two-year period. In total, 261 staff 2% of orders. However, medical staff required 15were surveyed, 32 were interviewed and there were seconds to prescribe a regular inpatient drug beforefive focus groups. The system led to 22 types of and 39 seconds after introducing the system.96risks of medication errors such as not allowing acoherent view of patients’ medications, mistaking Primary carepharmacy inventory displays for dosage guidelines, Research about e-prescribing outside hospital is lessplacing alerts on paper charts rather than in the frequent and sometimes less positive, though this issystem, separating functions that facilitate double standard in UK primary care.dosing and incompatible orders and generating A review of e-prescribing in outpatient settingsincorrect orders due to inflexible ordering formats. included 30 studies. Only one study found reducedThese risks occurred frequently.92 prescribing errors. There were no impacts onAnother US hospital implemented a commercially adverse drug events. Three studies found reducedavailable e-prescribing system to help reduce medication costs but five others did not.97mortality among children transported for Another study examined the impact ofspecialised care. Before and after analysis found e-prescribing in four US primary care practices.that the tool was associated with increased rates of There was no difference between those who usedmortality, not reductions.93 basic computerised prescribing and those usingIt may take some time for the benefits of handwritten prescriptions.98e-prescribing systems to become apparent and However some benefits have been observed.there may be difficulties in the transition or An analysis of 10,172 prescriptions in primaryimplementation period. An analysis of US data care found that a basic e-prescribing system wasfound that changing from using older e-prescribing associated with reduced medication errors.99to newer systems was associated with a reduction inprescribing errors from 36% to 12%. Improvements Compared to when using handwritten orders, thewere mainly a result of reducing inappropriate proportion of errors reduced from 18% to 8% in community-based US primary care. The largest improvements were in illegibility, inappropriate abbreviations and missing information.100THE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 18
  19. 19. But when three primary care clinics in the US It may be that decision support is more useful atimplemented e-prescribing, a time motion study some stages of the prescribing process than others.found that it took longer than handwritten A systematic review of 56 studies found that duringprescriptions.101 treatment initiation, decision support systems were more effective after drug selection, rather4.2 Decision support than before. Decision support systems were more effective in hospital than ambulatory settings andDecision support tools provide prompts to help when decision support was initiated automaticallyprescribers avoid errors when writing or entering by the system as opposed to the user. Combiningprescriptions. This subsection focuses on decision decision support with other strategies such assupport tools or alert systems that are standalone education was no more effective than decisionsystems (not part of e-prescribing) or where alert support alone.105systems are embedded in e-prescribing tools buttheir effects have been analysed separately. A Cochrane Review of 23 studies examined whether computerised advice about drug dosageHospital care improved processes or outcomes. ComputerisedEvidence about the benefits of decision support advice improved doses, reduced time to therapeutictools, such as alerts and prompts for prescribers, stabilisation and reduced the length of hospital stay.is mixed. It had no effect on adverse reactions. There was no evidence that integration into an e-prescribingA systematic review of computerised drug alerts system optimised effects. Interventions usuallyand prompts found that 23 out of 27 studies targeted doctors, but a few attempted to influencesuggested improved prescribing behaviour or prescribing by pharmacists and nurses.106reduced error rates. The impact varied based onthe type of decision support. Five out of 27 studies Often, decision support is an adjunct toreported benefits for clinical and health service e-prescribing. A paediatric intensive care unitmanagement outcomes.102 in Israel tested e-prescribing with or without clinical decision support. The rate of prescriptionAnother systematic review reported that four errors was 2.5% without any tools and 2.4%out of seven studies about standalone clinical once e-prescribing was introduced. There was adecision support systems found improvements in significant reduction to less than 1% when decisionmedication errors and three did not. Most studies support was added. E-prescribing decreasedwere not powered to detect differences in adverse prescription errors only to a small extent, butdrug events and evaluated small ‘home grown’ adding a decision support system had moresystems rather than commercial systems.103 impact.107A review of 87 trials of medication management A US trial tested the effectiveness of computer-information technology found that most trials: assisted decision support in reducing potentially inappropriate prescribing for older adults in A&E.–– focused on clinical decision support and 63 doctors using e-prescribing were randomly e-prescribing systems assigned to receive, or not to receive, decision–– took place in US hospitals support that advised against use of nine potentially–– focused on doctors inappropriate medications and recommended safer substitutes. The decision support group–– studied process changes related to prescribing prescribed one or more inappropriate medications and monitoring medication. during 3% of A&E visits by older people comparedProcesses of care improved for prescribing and with 4% of visits managed by those not receivingmonitoring in hospitals. There were few studies decision support. This was a statistically significantmeasuring clinical outcomes and these tended to difference.108show limited improvements.104THE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 19
  20. 20. Prescribing excessive doses is a common Similarly, researchers in the US tested whether aprescription error and can lead to adverse drug computerised alert system would reduce the rate ofreactions. In Germany, a clinical decision support errors in drug selection or dosing for people withsystem was tested that provided alerts about upper renal insufficiency. A total of 32,917 people weredose limits personalised to individual patient randomly assigned to usual care or the interventioncharacteristics. Before the system was introduced group, where a computerised tool was used to alert5% of prescriptions exceeded upper dose limits. pharmacists at the time of dispensing to possibleAfterwards, the rate of excessive doses reduced to errors in target drug selection and dosing. Of4%, with 20% less excessive doses compared with these, 6,125 people were prescribed one or morebaseline.109 of the target drugs over a 15-month period. Alerts helped to reduce medication errors. 33% of theA hospital in the US used decision support tools to intervention group and 49% of the usual care groupmeet the unique prescribing needs of children. An had medication errors at follow up.114advanced dosing model was designed to interactwith an e-prescribing system to provide decision While alerts can work well to reduce prescribingsupport for complex dose calculations for children. errors during the prescribing process or afterThe system was flexible and could be altered over prescribing, ensuring that prescribers ortime. It was well used and found to be feasible.110 pharmacists see alerts may be an issue. Researchers in Australia tested whether decision supportOther researchers in the US examined decision within a hospital e-prescribing system influencedsupport alerts for helping avoid errors when putting medication ordering on ward rounds. 46 doctorsmedication orders into an e-prescribing system. were shadowed during ward rounds and 16 wereData for all patients at five community hospitals interviewed. Senior doctors influenced prescribingover a six-month period were analysed. The alert decisions during ward rounds but rarely used thesystem changed doctor’s behaviour and patient e-prescribing and alerts system. Junior doctorstherapy 42% of the time and reduced medication entered most medication orders into the system,errors.111 often ignored computerised alerts and never raisedAs with more generic e-prescribing systems, their occurrence with other doctors on warddecision support tools have also been used to rounds. Doctors did not think that most features ofidentify potential errors after prescribing has the decision support system were useful.115occurred. A hospital in Japan tested an alert system Alerts are not the only type of decision supportfor evaluating kidney function and checking system. Decision support tools may also includedoses of medication according to the patient’s access to clinical information and guidelines.renal function. Discontinuation of inappropriate Researchers in France tested whether makingmedication for those with poor renal function guidelines about antibiotics more accessible torose from 24% to 54% after the alert system was doctors would increase adherence to guidelines. Inimplemented.112 this instance, a lack of adherence was specificallyAlerts targeting pharmacists have also been tested. defined as a prescribing error. One hospitalThese focus on identifying errors once prescriptions changed from having guidelines available inhave been entered. In the US a computerised tool booklet format on wards to embedding thesealerted pharmacists when people aged 65 and older guidelines into an e-prescribing system. Assessmentwere newly prescribed potentially inappropriate of 471 consecutive antibiotic orders for pneumoniamedications. In total, 59,680 older people were before and after the change found improvementsrandomised to intervention or usual care groups. in the daily dose and the planned duration ofAlerts helped to reduce inappropriate prescriptions treatment.116for two drugs.113 In the US, a computerised guideline increased use of appropriate medication and decreased errors in drug doses.117THE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 20
  21. 21. Other researchers in the US examined three 4.3 Human factors issuespersonal digital assistant (PDA)-based drug Few studies have examined how healthinformation sources for reducing potential professionals interact with e-prescribing andmedication errors. All three PDA tools were decision support systems and the human factorsfound to be feasible and one was found to be more issues that may be influential. But there is someeffective than the others.118 evidence of scope for further work in this area.Primary care Implementation factorsLittle has been written about standalone decision E-prescribing systems are common in the UK.support or alert systems for reducing errors in This contrasts with the US, where the use ofprimary care. The evidence that does exist tends to e-prescribing systems has been strongly advisedbe mixed. nationally, but rates of adoption remain relativelyThe US Food and Drug Administration (FDA) low. Eight focus groups in US primary care foundissues black box warnings about medications with that e-prescribing was thought to improve theserious risks. Doctor adherence to these warnings availability of clinical information, prescribingis low. A system was tested for inserting black box efficiencies, coordinated care and documentation,warning alerts about drug-drug, drug-disease and and result in safer care. Factors supportingdrug-laboratory interactions into an outpatient adoption included human factors features suchelectronic health record with clinical decision as organisational support, adequate time, a shiftsupport. The alerts did not increase adherence to in staff workload, equipment stability, educationthe black box warnings.119 about changes in patient interactions and positive attitudes.122On the other hand, following implementation ofalerts cautioning against prescribing certain drugs In another part of the US, a community basedto elderly people in some US outpatient clinics, integrated health system implemented athere was a 22% reduction in exposure of elderly computerised order entry system. Strategiespatients to these drugs.120 for successful adoption included senior buy-in, ongoing communication, a team-oriented culture,A review of computer decision support for iterative implementation, ongoing readily accessibleimproving prescribing in older adults in primary training, gaining buy-in from clinicians andcare or hospital included 10 studies. Eight of these workflow redesign.123studies found some improvement in prescribingincluding minimising drugs to avoid, optimising Workflow redesign is gaining more attention,drug dosage or improving prescribing choices. Few but knowledge in this area remains limited.studies reported clinical outcomes.121 Researchers in the Netherlands tested the effects of an e-prescribing system on inter-professional workflow. In total, 23 doctors, nurses and pharmacists at one hospital were interviewed and documents were reviewed. The system reorganised existing work procedures and impacted on workflow in positive and negative ways. It reassigned tasks and areas of expertise and fragmented patients’ medication-related information, while providing limited support for professional groups to coordinate their tasks.124THE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 21
  22. 22. Three sites in the US implementing e-prescribing Types of alertsidentified barriers, including those relating to The effectiveness of e-prescribing systems andhuman factors. Implementation barriers included decision support may sometimes be modestprevious negative experiences with technology, because clinicians often override electronic alerts.initial and long-term cost, lost productivity, Two US teaching hospitals tested an alert that didcompeting priorities, change management issues, or did not allow the information for a certain drugfunctional limitations, IT requirements, waiting combination to be entered on the system. Of thosefor an ‘all in one’ solution and confusion about in the intervention group, 57% did not reorder thecompeting systems.125 Another study identified 15 alert-triggering drug within 10 minutes of receivingbarriers to using medication alerts at five primary an alert compared to 14% in the control group.care clinics in the US.126 In other words, prohibiting input of some drug combinations reduced errors of this type. However,Design features unintended consequences included serious delaysHuman factors approaches are concerned with in treatment.130how technologies are designed to be most useful The impact of active versus passive alerts, alerts thatand user friendly. A systematic review of 19 pop up versus those that are just inserted into thestudies examined the impact of design aspects of online record and alerts that require the prescribere-prescribing systems on usability, workflow and to acknowledge reading them have all been tested.prescriptions. 16 studies were qualitative and three In the US, alerts were built into an e-prescribingused mixed qualitative and quantitative methods. system to help doctors take account of changingDesign aspects were found to be important kidney function when prescribing medications.for increasing use of the systems and reducing When treating 1,598 hospital patients with acuteprescribing errors. Such design aspects were kidney injury, doctors received passive non-categorised into seven groups: documentation interactive warnings from the e-prescribing systemand data entry components, alerts, visual clues and on printed ward round reports. An interruptiveand icons, dropdown lists and menus, safeguards, alert was provided for contraindicated or highscreen displays and auxiliary functions.127 toxicity medications that should be avoided orAnother review of 41 randomised trials adjusted. This alert asked prescribers to modify orexamined whether design features of prescribing discontinue the orders, mark the dosing as correctdecision support systems predict successful or defer the alert to reappear next time. The activeimplementation and usage. 37 studies reported alerts were associated with more modifications orsuccessful implementation, 25 reported discontinuations and more prompt action. Passivechanging professionals’ behaviour and five found alerts had limited response.131improvements in patient outcomes. No design Decision support tools may generate large numbersfeature was more prevalent in successful trials.128 of insignificant on-screen alerts presented as pop-upCognitive fit between the user interface and boxes. This may interrupt clinicians and limit theclinical task may impact on whether doctors use effectiveness of these systems. A randomised trial ine-prescribing systems. Cognitive task analysis of England compared the impact of pop-up and non-clinical alerts for antibiotic prescribing in a US pop-up alerts on prescribing error rates. 24 juniorneonatal intensive care unit found that responses doctors, each performing 30 simulated prescribingto alerts may be context specific and that a lack of tasks in random order, were shown pop-up alerts,screen cues increases the cognitive effort required non-pop-up alerts or no alerts. Doctors receivingto use a system.129 pop-up alerts were about 12 times less likely to make a prescribing error than those not shown an alert. Doctors shown a non-pop-up alert were about three times less likely to make a prescribing error than those not shown an alert.132THE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 22
  23. 23. Similarly, researchers in the US aimed to improve 4.4 Standardisedclinician acceptance of drug alerts in 31 primarycare practices by prioritising alerts in order to medication chartsreduce workflow disruptions. Over a six-month Other tools to support the interface between healthperiod, 71% of alerts were non-interruptive professionals and the systems and environmentsand 29% were interruptive. Two thirds of the in which they work have been researched in lessinterruptive alerts were accepted.133 depth, but some studies are available.The majority of prescribing alerts may be ignored Computerised medication charts have been tested.because they are not seen as clinically relevant. In a system very different to that used in the UK, aBeing able to customise when alerts are seen may hospital in the Netherlands compared a medicationincrease their usefulness. A Canadian study tested distribution system where the transcription oftwo approaches to medication alert customisation: handwritten into printed medication orders takeson-demand versus computer-triggered decision three to five days versus a computerised medicationsupport. Doctors randomised to on-demand alerts chart which was updated daily by pharmacyactivated the drug review when they considered assistants on the ward. The prescription error rateit clinically relevant. Doctors randomised to was higher with computerised charts (50% versuscomputer-triggered decision support viewed all 20%) but this was due to more administrativealerts for electronic prescriptions in accordance errors, such as omitting the prescriber’s name andwith the severity level they selected. Customisation the date. The rate of errors with potential clinicalof computer-triggered alert systems was more significance was lower because duplicate therapyuseful in detecting prescribing problems than on- was eliminated.137demand review. There was no difference between In Australia, a standard medication chartgroups in prescribing errors. The majority of alerts was developed for recording prescribing andwere ignored because the benefit was judged greater administration of medication in hospital. Beforethan the risk.134 and after audits in five sites found the prescribingResearchers in the US tested alerts that required a error rate decreased from 20% of orders per patientresponse from doctors to prevent concurrent orders to 16%.138of warfarin and non-steroidal anti-inflammatory After preliminary testing, the standardiseddrugs. In total, 1,963 doctors were assigned to medication chart was rolled out to 22 Australianreceive passive alerts or active alerts which required hospitals. Prescribers were educated and baselinea response. Active alerts had no benefits over audit findings were presented when the chart waspassive alerts.135 introduced. Prescribing errors decreased by almost one third.139WorkforceA hospital in England tested computerisedprescribing with alerts over a three-month period. 4.5 Other toolsSenior doctors and those more experienced using A number of computerised and other tools havethe system were more likely to ignore a warning been tested to reduce prescribing errors, often inmessage.136 conjunction with electronic prescribing. These interventions are a mix of tools to reduce errors during prescribing and tools to identify and mitigate errors before they reach the patient. One study examined the effect of regular and expected printed educational materials on prescribing. In Canada, 499 doctors wereTHE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 23
  24. 24. randomised to receive 12 evidence-based drug receives from different hospitals nationwide.therapy letters immediately or after 3–8 months This system was used to address the problem of(control group). The aim was not merely to improve duplicate medications for outpatients visitingevidence-based prescribing, but also to reduce multiple hospitals. At one hospital an e-prescribingdosage and drug choice errors. The series of letters system was enhanced with the ability to accessinfluenced which drugs were prescribed to newly smart cards and alert doctors about potentialtreated patients. Each letter alone did not make a duplicate medications at the time of prescribing.significant impact, but when combined they made a Over a three-month period, 2% of all smart cardsdifference.140 read contained medications that would potentially have been duplicated without this system. AroundA hospital in the US introduced a voluntary one-third of these prescriptions were revised due tointeractive computerised worksheet for use when the alerts.146prescribing parenteral nutrition in the neonatalintensive care unit. The worksheet reduced the Combining more than one tool is becomingprescribing error rate from 14% to 7%.141 popular. A US hospital system implemented a range of clinical information technology suchAnother US hospital tested a standardised as e-prescribing, pharmacy and laboratorychemotherapy order form to reduce prescribing information systems, clinical decision supporterrors and the cost of medication to reduce systems, electronic drug dispensing systems and avomiting and nausea. The form was associated with barcode point-of-care medication administrationfewer prescribing errors and a reduction in the system. Medication errors decreased. Mostaverage cost.142 prescribing errors decreased, including drug allergyAnother US hospital examined the impact of detection, excessive dosing and incomplete oradding a medication list targeting the most unclear orders.147common medications to an e-prescribing system ina paediatric A&E department. The medication listdecreased errors from 24 to 13 per 100 visits.143Elsewhere in the US, a hospital tested a system forreconciling medications that patients take at homewith what they receive in hospital. The unintendeddiscrepancy rate between a patient’s homemedications and admission medication orderswas reduced from 20% to 1% using the electronicreconciliation system.144A hospital in Sweden tested providing amedication report for older people dischargedinto the community. 32% had one or moremedication errors compared to 66% of aretrospective comparison group who did notreceive a medication report. Prescribing errorswere not identified separately.145In China and Japan, patients may ‘shop around’ fordoctors or hospitals, visiting a number of doctorsfor the same condition. In Taiwan, a nationalinsurance health smart card was adopted, whichcarries information about the medications a patientTHE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 24
  25. 25. 5. Summary5.1 Key points errors, particularly if the systems do not allow the prescriber to see the entire medication history orMost people taking medication will benefit from other relevant information easily.it, but there is always the potential for errors whichmay cause harm. Prescribing errors are the largest Alerts and prompts alone have generally not beensource of medication errors. A systematic review of found to reduce prescribing errors, though some16 studies about errors in handwritten prescriptions studies have positive results.in hospitals found that the most common causes oferror were mistakes due to inadequate knowledge Although opinion pieces and narrative articlesof the drug or the patient, memory lapses, lack are available,149 less empirical research has beenof training or experience, fatigue, stress, high published about ‘human factors’ approaches toworkload and inadequate communication between reducing prescribing errors regarding the interfacehealthcare professionals.148 between personnel and the environment and systems in which they work.A number of strategies have been tested to reduceprescribing errors. The most commonly researched Training staff to fulfil their roles is an importantstrategy involves redesigning equipment and human factors component. There is some evidencetasks through the use of electronic tools such as that training medical students can help them feele-prescribing and computerised decision support more confident about prescribing but the longer-systems (alerts and prompts). While a great deal term impact on reducing errors remains uncertain.has been written about e-prescribing and alert tools Other studies have examined training for fullyin hospital, and to a lesser extent in primary care, qualified doctors. This has taken the form of one-evidence about the effectiveness and value of such to-one sessions about specific medications orsystems is mixed. In the US e-prescribing systems patients (academic detailing), group sessions andhave been mandated for widespread use, while in collaborative improvement projects and qualitythe UK such tools are very common. Some research circles where groups of prescribers network, sharesupports this, with findings of substantial reductions good practice and take part in practical errorin prescribing errors. In fact, it is common for the reduction initiatives.introduction of combined e-prescribing and alertsystems to halve prescribing errors. Some research is available about expanding pharmacist roles to target error reduction,However, other studies suggest that the types of particularly in hospital. Research is also emergingerrors affected may be clinically insignificant and about pharmacist roles in primary care. Studiesthat there may be other costs involved. While have examined reactive use of pharmacist roles,e-prescribing systems reduce illegibility errors, such as using pharmacists to review prescriptionssuch systems may take more time than handwritten for errors before medication orders are filled.prescriptions and may introduce new types of Research is also emerging about more proactive use of pharmacist roles, such as circulating on wards to check prescriptions and providing education one to one or in groups to prescribers.THE HEALTH FOUNDATION Evidence scan: Reducing prescribing errors 25